We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.DC

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Distributed, Parallel, and Cluster Computing

Title: Decentralized Wireless Federated Learning with Differential Privacy

Abstract: This paper studies decentralized federated learning algorithms in wireless IoT networks. The traditional parameter server architecture for federated learning faces some problems such as low fault tolerance, large communication overhead and inaccessibility of private data. To solve these problems, we propose a Decentralized-Wireless-Federated-Learning algorithm called DWFL. The algorithm works in a system where the workers are organized in a peer-to-peer and server-less manner, and the workers exchange their privacy preserving data with the anolog transmission scheme over wireless channels in parallel. With rigorous analysis, we show that DWFL satisfies $(\epsilon,\delta)$-differential privacy and the privacy budget per worker scale as $\mathcal{O}(\frac{1}{\sqrt{N}})$, in contrast with the constant budget in the orthogonal transmission approach. Furthermore, DWFL converges at the same rate of $\sqrt{\frac{1}{TN}}$ as the best known centralized algorithm with a central parameter server. Extensive experiments demonstrate that our algorithm DWFL also performs well in real settings.
Comments: The proof is not perfect in section 4
Subjects: Distributed, Parallel, and Cluster Computing (cs.DC)
Cite as: arXiv:2109.09142 [cs.DC]
  (or arXiv:2109.09142v2 [cs.DC] for this version)

Submission history

From: Shuzhen Chen [view email]
[v1] Sun, 19 Sep 2021 15:11:42 GMT (7171kb,D)
[v2] Wed, 22 Sep 2021 09:22:35 GMT (0kb,I)

Link back to: arXiv, form interface, contact.